Using Regression Discontinuity Designs to Control for Selection Bias in Evaluations of Differential Item Functioning

• In Event: AERA Poster Session 14
  In Poster Session: Division D Section 1 Poster Session 1

Abstract

Traditional methods for detecting differential item functioning (DIF) do not control for selection bias, so differences in item functioning may be confounded by pre-existing group differences. The purpose of this study is to demonstrate the application and practicality of embedding regression-discontinuity (RD) designs into evaluations of DIF of academic achievement items in order to control for selection bias between DIF groups, where groups are defined by language of administration and a pre-test language screener determines language of administration. To carry out this demonstration, a secondary analysis of the Early Childhood Longitudinal Study, Kindergarten Class of 1998-99 (ECLS-K) was performed. Results indicate that conclusions can vary substantially depending on whether the method for detecting DIF controls for selection bias.

Authors

• Marc Goodrich, Texas A&M University - College Station

• Natalie A. Koziol, University of Nebraska - Lincoln